On Monday 27th January 2025, $1tn of market value was wiped from the US tech index thanks to the release of the Chinese AI chatbot DeepSeek.
The model slightly outperforms market leaders like OpenAI’s ChatGPT but the main reason it has been so disruptive is its cost – both to create and to use.
Microsoft is investing $10bn in OpenAI in a multi-year deal and it’s estimated that Google has invested $191bn in their AI product Gemini. [1] DeepSeek took 2 months and £6m to train {2] – a fraction of anything else on the market. This has allowed them to release a robust free app direct to consumers, but also open source the model under the MIT license – making it a highly appealing prospect for developers to build upon.
It’s easy to see why the financial markets were spooked. I think the more interesting thing to examine is what it means for the AI industry in terms of what it can technically deliver. Tools from the big players like Copilot, Gemini, and Apple Intelligence have promised solutions to problems that don’t really exist like re-writing your emails or summarising documents that you should really be reading anyway as part of your job. This kind of “solution without a problem” style of AI development isn’t going to cut it anymore for two reasons.
The first is that companies that have been wrapping other models to build their products now have a more appealing model to wrap. Obviously, this has cost benefits but also allows the product build to focus on UI/UX. At SAMY, we have two products (Brand Guardian and SAMY Assistant) that fit this bill. They wrap Chat-GPT and apply some prompt engineering with our own theoretical frameworks and mental models to tailor the output for users. Because the cost to build is so much lower, we can spend our time designing experiments for better prompt engineering, or A/B testing UI changes, and be more agile in responding to user feedback in general.
The second is that if a company did want to enter the space and build an LLM with their own weights trained for hyper-specific use cases, they can. The barrier to entry has now been demonstrated to be far lower than previously envisaged – which is why NVIDIA took the brunt of the market drop, losing $600bn in value. Their chips aren’t the only route to market anymore.
The moat has been drained and the water is used to make a delicious smoothie. But more than that, there’s an opportunity for companies to listen to consumer needs and redefine what an AI castle fundamentally looks like. If you are interested in how to engage with AI in a mindful way, get in touch with us.
As SAMY Alliance continues to grow through strategic acquisitions and organic expansion, our vision remains clear: becoming a global independent leader in the Social Media Marketing space leveraging tech, data, and intelligence-fueled creativity capabilities and facing the future to be able to make the boldest brands matter.
[1] https://fortune.com/2024/04/18/google-gemini-cost-191-million-to-train-stanford-university-report-estimates/?utm_source=chatgpt.com
[2] https://www.reuters.com/technology/artificial-intelligence/big-tech-faces-heat-chinas-deepseek-sows-doubts-billion-dollar-spending-2025-01-27/?utm_source=chatgpt.com
This analysis was done by SAMY’s Data Director, Mike Tapp.
